摘要
针对临床尿沉渣显微图像识别问题,提出了一种新的图像预处理方法和基于细胞神经网络(CNN)的分割算法.该方法通过拉伸图像中各个像素的灰度值与局部灰度值之间的差来增强图像中目标的边界,通过对局部灰度均值的非线性变换来消除图像中光照的不均匀,进而设计出合适的CNN模板来分割图像,最终利用形态学操作得到分割结果.通过对100幅临床尿液样本图像的测试,并与传统的阈值分割法相比,该方法获得了更加连续的边界和更加准确的目标分割结果,并已集成到全自动尿液粒子分析系统中,应用于临床,取得了良好的效果.
In view of urinary microscopic image recognition, a novel urinary sediment image segmentation approach based on cellular neural network (CNN) was presented. Before the image segmentation, a preprocessing scheme was performed, which enhances the edges of objects in the image by stretching the difference between every pixel gray value and the local gray mean value, and eliminates the disequilibrium of illumination by nonlinear transform of the local gray mean value. Suitable CNN template was then designed to segment the images and morphological methods were finally carried out to get final results. The experimental results with 100 clinical urinary images showed that this approach provides better boundaries and more accurate object detection campared to the conventional threshold segmentation. The related algorithms were successfully integrated into automatic urinalysis systems, and gained good results in clinical applications.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2008年第12期2139-2144,共6页
Journal of Zhejiang University:Engineering Science
基金
国家自然科学基金资助项目(60772092)